Rock bands, choirs and symphony orchestras each pose their own challenges to the engineer. Put them all in the same room, and things get really messy...
The success of their fourth album, The Seldom Seen Kid, has elevated Elbow from indie hopefuls to national treasures. Back in December last year we interviewed band member and producer Craig Potter about the recording of the album itself, and since then, the material has continued to evolve. Its most ambitious outing has been Live At Abbey Road, in which the band performed the entire album live in Studio 1 at London's most famous studio for BBC Radio 2, with the backing of the BBC Concert Orchestra and and Chantage, Radio 3's Choir of the Year.
The man charged with overseeing and mixing the live recording was BBC staff engineer Rupert Flindt. He was aware from the outset that the project would be challenging. "I'd just been in Abbey Road 1 rehearsing songs with the Brit School and orchestra. We'd found that as soon as anyone played the drums, even on stuff from musicals, you couldn't hear a thing. It's a room that's utterly designed for classical music. It's got a very rich 300‑400 Hz hump that becomes a 'boom' if you do something very loud in there — but actually lends itself to giving strings a rich sound, hence all the film soundtracks they do there.”
As minds were changed further up the chain of command, moreover, the potential for disaster grew. "At first I was told it was going to be a studio recording, and they have a big drum and rhythm‑section booth, so we thought we'd put the drums and guitar cabs in there, and the band out [in the main studio] on in‑ears. Then the next conversation I had, they said they were going to have an audience of 200 in there to watch the thing. Which immediately meant that the drums had to come out into the room and all the layout had to be much more performance‑orientated, and we had to have a PA there for the audience. So that was quite a large rethink right at the beginning!”
Key to the success of the project was the band's own understanding of the issues. "I had some emails with Craig, who's the producer, and the thing that made me realise it would be possible was when he wrote back and said, 'We will play within the dynamic of the orchestra. We know about strings and it won't work if we play as a proper rock band, so we'll wear in‑ears and try to contain the volume on the guitar cabs.' Another bonus was the recording experience that conductor Mike Dixon brought to the show. We'd worked together before and I knew he was sympathetic to the recording process and really good at holding such differing talents together.
"And then,” adds Flindt with a wry smile, "the choir came in on the equation! Choirs, like violins, are very difficult to do when there's a lot of noise going on. So we had a full string section of 34 players and a chorus, all of whom were going to need to be heard against a rock band in a lively room — with an audience and a PA as well.”
As if setting up and recording such a performance was not enough of a challenge, budget and time constraints meant that even though it was not a live to air broadcast, Flindt would have to mix it live to stereo — a far from trivial task, with almost 100 signals coming into Abbey Road's control room. "The budget to remix for the radio and TV broadcast wouldn't be there, but at the same time the band wanted to make a CD and DVD of it, so they wanted a multitrack made, which worked to both our advantages. Because it was over two days — a rehearsal Friday afternoon, then a rehearsal Saturday afternoon and the show Saturday night — we could test out the multitrack on the first day, and then on Saturday morning I had the multitrack playbacks linearly in the signal path, so I could just spin the multitrack and actually rehearse the songs for a couple of hours that morning.”
With only two days to get in, set up, rehearse and record, careful planning was, naturally, essential. "The band line‑up went to about 40 channels, which in itself is not abnormal for a band like Elbow, but I knew I was also going to have to put an orchestra on top of that, and I knew the only way I was going to get the orchestra to work was by close‑miking, and that means a mic on every instrument. I then had to find out every player in the orchestra, assign a microphone to them, and add up the tracks.”
To keep things a little more manageable, Flindt took a chance and asked the chorus singers to share mics in pairs. "If I was doing a gospel choir or something like that, you would always give them a mic each, but I took a slight risk because of the kind of sound, and also we just didn't have enough tracks. With everything as it stood, if I'd miked the chorus it would have come to 127 lines for everything, and the band wanted the multitrack to have every instrument. So we were already in a log-jam to get it down to 96 tracks, which is probably the largest viable multitrack format — the band wanted it to be done on Pro Tools at 24‑bit, 96kHz.”
Dealing with 34 individual string mics at the mix would also have been a bridge too far, so Flindt delegated this very specialist task to colleague Gary Parker. "We've developed a technique when we do rock bands with strings. We hire a whole set of DPA 4060 little omni mics, and they come with a specialised triangular clip that goes onto the bridge, and the orchestra are aware of these and prepared to let us touch the violins — which is not always the case with orchestras! So every violinist had a little omni mic clipped to the bridge, and the four double‑basses had [Neumann] U47 valves, one on each bass — because we could!
"Gary Parker stays out in the house with a Midas Verona [desk], which he's very used to, and he takes all those bugs into the Verona. He knows the EQs that work — if you put your ear next to a violin [to hear what a clip‑on mic hears] it sounds awful, a very shrill, boxy, sound — and we've developed a really nice way of EQ'ing them and subbing them down to 10 stereo groups. You suck out a couple of mid frequencies and the boxiness, putting in some of the 2‑300Hz weight — which is what the studio does if you're doing it acoustically — and he EQs the first and seconds one way, the violas another way and the cellos a slightly different way. What I end up getting is 10 lines from him — stereo firsts, stereo seconds, stereo violas, stereo cellos, stereo basses — and I know that at that point, what I'm getting is a very nice string sound, given that it's bugs. Then I can very slightly EQ them into context.
"The other problem with the bugs is that they've got little connectors on, so they tend to crackle and rattle, and he's out in the house with headphones continually listening through to them, so if there's a splat or a crackle, he can instantly mute that mic, wander out and fix it.”
Another difficult question was how to place all the musicians within Studio One. "It's a big room — but it's not a big room once you've put in a symphony orchestra and a chorus and a rock band and a PA and monitors and a crowd of 200. And then there are the fire exits! It becomes the art of arranging the orchestra physically so that the bleed is as minimal as it can be. If seven or eight brass players are sat behind the strings in a classical style, their sound pours into the string mics. That's fine if it's a classical piece, but with this stuff you need to try to get separation between the sections. So we placed the brass in front of the strings on the left, not behind them; we placed the winds to the right of the strings, we pulled the percussion away from the strings to the stereo right, and then we put the chorus right at the back so that the main spill on the chorus was reflection off the back wall.”
Careful attention to mic placement and polar pattern was also vital. "Apart from the fiddles, we were using directional mics all the way through — not very tight cardioids, but orienting the mics in such a way that we bring the wind mics in so that the back of their cardioids is where the loud brass is. It's not just what the mic points at — there's an element of what the mic doesn't point at, as well. One of the differences between cheap cardioid mics and good ones isn't how they sound at the front, it's how they sound off‑axis, and as they get better and better, the off‑axis sound gets sweeter.
"There are also big health and safety issues nowadays about sound levels. So you can't have players in front of other players because of noise exposure, and you have limited rehearsal times, and you can't really have a rock band in the same room, fundamentally! It really helped that the band were on in‑ears, and the guitar amps were just loud enough to be heard and sound good, and I screened them into half‑boxes so they were mainly firing into the audience.
"We rigged the stage, we rigged the drums, and then we brought in the traditional four‑screen drum tent. Unfortunately, the drum kit was about three foot too big in all directions, and however we tried to orientate the drum box, it wouldn't go. So eventually we had to build a circular drum room out of lots of screens, and cover the roof with some very heavy fabric to stop things going out of the top — which actually worked quite well because it didn't become a boxy chamber, so the drum sound was better for it, and the absorbent top meant I could put high overheads up and not get reflections off that. In the end that brought the level down just enough, with Jupp playing quite tightly as well.
"I had a pair of audience mics on each side, which doubled up as the ambient feeds to the in‑ears for the band, so they got some sense of what was coming off the audience. And those were mounted as close as possible to the PA speakers, which is a little counter‑intuitive, but the reason is that as soon as you move them away from the PA you start to introduce delay into it, so they become more and more unpleasant and spilly — whereas if you put them as close as you can beside or on top, the spill you get from the PA into them is very coherent with the direct sound, it's not got that comb‑filter or slap distance, and you go off‑axis for level. I actually put a pair of [Shure] 57s in, which are good old tight vocal mics, pointing right down at the front of the crowd, and then a pair of [Neumann] KM84s, which have got a nice smooth open sound. You have to do quite a bit of EQ to tighten up audience mics, but as soon as you're being filmed, it's really important to have the audience in there — you've got to have the cheering and the whistling and the shouting and the singing along.
"Having a PA was the real fly in the ointment for us in that room: we had to have it, and the audience had to have an experience, so [hire company] Richard Knowle produced a very focused PA of mostly small Meyer UPJ speakers arranged across the front, angled right down on the crowd. Elbow provided a front‑of‑house guy that they like, who knows their stuff, and with Richard Knowle came a guy named Tony Birch who I've worked with a lot and understands my problems as a broadcast engineer relative to the audience. He is great at steering things just loud enough for them, just quiet enough for me, and taking out some of the troublesome spill frequencies from the front‑of‑house.”
In order to capture a pristine multitrack recording for future remix and CD/DVD release, the live mix was actually done using the signals coming through the Pro Tools rig. "All the Pro Tools was there for was to get the cleanest possible version of the mics at the highest quality. So we used the Neve 88 mic pres at the top of the desk, plus a rack of another 24, to feed the Prism A‑Ds straight into Pro Tools at line level, with Abbey Road's Sam Okell, who was running the Pro Tools with me, keeping an eye on those all the way through the rehearsal, to get the levels. I know from experience that orchestral brass will go up at least 10dB on the performance, with a couple of (soft?) drinks and a red light and an audience! No plug‑ins in the Pro Tools were needed. We went out of the Pro Tools at line level on the same D‑A and back into the Neve 88 at line level into its monitor path, and then all my EQ and all my compression and all my balance was done in that monitor path. It meant that I could fiddle around as much as I wanted without interfering with the multitrack at all. I think it ended up on 88 tracks, but we also multitracked my string mixes and my wind mixes, in case the band didn't have time later to mix it in detail.”
The arrangement had the advantage that some submixing could be done within Pro Tools in order to get the 88 recorded tracks down to the 72 desk channels available on Abbey Road's Neve 88VR. However, it also meant that "Everything we did was through the latency of the Pro Tools, including the click for the orchestra and choir's headphones. Nobody in the entire setup said anything about any problem, though, because everything went through the same very quick path. Phasing always comes when you start mixing direct with latent [signals].” The exception was the band's in‑ear feeds from their own instruments, where phasing with the orchestral instruments was not much of a problem because the physical distance between band and orchestral mics was so great.
Even with Gary Parker's help, and the 100+ sound sources condensed to 72, the prospect of mixing the entire show live to stereo was a daunting one; and, again, careful preparation was vital, as was the Neve desk's snapshot automation. "You've got to know your album from beginning to end, and at the end of the first number, be ready with your faders for the second one,” explains Rupert Flindt. "I felt the only way to do this was to use every bit of technology we could, but we couldn't automate the mix, because it was happening live. So we used the rehearsal multitrack we did on the Saturday morning, and the Saturday afternoon rehearsal with the orchestra. We recorded those, and then went to the beginning of each number and set the starting balance for each one, so I knew the relative levels of the band internally, and whichever orchestral instruments were in were faded up. But all the way through this I was trying to keep the orchestra down when they weren't playing, in order to keep some degree of coherence and tightness to the sound. I would always try to start each track with the minimum amount of orchestra I needed, and then through the track I could open it out gently when the parts came in, so you didn't get jarring transitions — if I was to suddenly open all the string mics when they come in, the whole thing would suddenly go 'booom', whereas I knew that if they were coming in in eight bars, I could just ride them up through the bit before.
"So we pre‑snapped the starting point of each track, but not the audience mics and not the vocals, so the vocals and audience were isolated from the snapshot recall. What would happen is that at the end of the first track, the audience did their cheering, so I pushed the audience pretty high, to hide the fact that I then scrolled through the list on the Neve snapshots and recalled the snap for the beginning of the second track. I had to have the vocals separate in case Guy was talking, so I couldn't snap the vocal positions; also, someone speaking is a hell of a different gain from someone singing, and there's always quite a battle if you have singers who chat.
"Danny [Evans, Elbow's live sound engineer] sat with me in the cubicle as well, because he knew the arrangements of the songs in more detail, so he could say, 'Guitar coming here, watch out for the keyboard there, important bit of percussion there...'. He sat on my right, and on my left, I had Al Booth, the BBC producer, with Nick Ingham's wonderful orchestral scores. And I laid the desk out so all the orchestra were down the left side of the board, the band were down the right side, and I was in the middle with the control group masters, listening to Al calling wind parts, or harp, or whatever, and then Danny on my right saying, 'Don't need so much guitar here,' or 'Synth bass here,' and doing the de‑essing for the vocal on his side. So it was really, in a way, a three‑person job to steer it all the way through — not counting in‑ear monitors and front of house.
"The other thing we did was run the fader automation on the desk through the actual show, sync'ed in timecode to the multitrack with time of day. I didn't know at this point how much I was going to f**k up — I could easily have not had a vocal up, or missed something important — so we recorded all my fader moves, which helped us in two contexts. One was in 'Some Riot'. Guy sings 'f**kers', which is not a BBC word! So at the end of the concert we had to take the multitrack back to that point and cut his vocal, and that just became a five‑second patch. Also, at the end of 'One Day Like This' I overcooked the choir a bit, so we came back in and repaired the last 16 bars, and there was some slightly over‑excited brass playing, so we buttoned out one brass player. Those were the only two actual repairs to the stereo mix that we did live, but we had the option if we had time, to repair a lot, if it had gone wrong.”
Fortunately, nothing did go wrong, and two days of ridiculously hard work were rewarded with a luxury not usually afforded on BBC projects: the chance to relax with a well‑deserved glass of red wine "while Abbey Road employees de‑rigged everything for us!” The show was broadcast in January this year to wild acclaim, getting the highest ever Red Button audience figures for a music session, and a CD/DVD package is now available. Rupert Flindt is quick to point out that for all his efforts, it's the music that matters above all: "There's an old sound engineer's proverb: How do you get a great guitar sound? Put any mic in front of a great guitarist! Elbow are a great band.” .
In conventional multitrack recording, we're accustomed to thinking of compression as something that can make the mix engineer's life easier. Unfortunately, as Rupert Flindt explains, this doesn't apply in a situation where there are more than 100 mics open in the same room! "Once you get a load of multiple mics out, compression becomes a real beast, because it alters spill continually. With a multitracked rock recording, you can compress things perfectly for each track, but the minute you've got two compressors going in the same room on two mics, when this guy plays his level's there — fine — he stops playing, whoop, up goes his gain and the spill floods across, so all sorts of weirdnesses happen. You'll find this if you're doing a vocal and a guitar: if you start putting in your standard vocal compression, when you stop singing, the guitar sound goes weird. So in this show there was a limited amount of compression. I used Al Smart C2 on kick and snare, mono on each, I used Urei 1176s on the vocals, and a little bit of desk compression on the keyboard parts — really as a safety net, because live keyboard players can be unpredictable.
"Then you have to start thinking about global compression. At various points during the day, we used an SSL Quad Bus 1U compressor, we used a Neve 33609, and everything was very difficult to set up because what was coming at us was so unpredictable and so live. So we ended up using a tiny, tiny bit of SSL compressor across the whole thing for the TV recording, just so when stuff shot through there was something to hold onto it, but we left the whole thing uncompressed underneath that. Because it wasn't live to air, it was just live to recording, I took away a pretty much uncompressed stereo mix, put it through Pro Tools and did a much more careful limiting job on it, along with a tiny bit of EQ because of the unfamiliar monitoring. It didn't need very much. The plug‑in used, which I love, is the Massey L2007, which is very good value and seems to have a really nicely crafted limiter. So I put it through that and got it all to BBC broadcast levels, which means peaks are around ‑10dBFS, or what we call PPM6.”